Search

Descriptive fit indices that do not require a formal statistical basis and do not specifically depend on a given estimation criterion are useful as auxiliary devices for judging the appropriateness of unrestricted or exploratory factor analytical (UFA) solutions, when the problem is to decide the most appropriate number of common factors. While overall indices of this type are well known in UFA applications, especially those intended for item analysis, difference indices are much more scarce. Recently, Raykov and collaborators proposed a family of effect-size-type descriptive difference indices that are promising for UFA applications. As a starting point, we considered the simplest measure of this family, which (a) can be viewed as absolute and (b) from which only tentative cutoffs and reference values have been provided so far. In this situation, this article has three aims. The first is to propose a relative version of Raykov's effect-size measure, intended to be used as a complement of the original measure, in which the increase in explained common variance is related to the overall prior estimated amount of common factor variance. The second is to establish reference values for both indices in item-analysis scenarios using simulation. And the third aim (instrumental) is to implement the proposal in both R language and a well-known non-commercial factor analysis program. The functioning and usefulness of the proposal is illustrated using an existing empirical dataset.

A Monte Carlo simulation study was conducted to compare fit indices used for detecting the correct latent class in three dichotomous mixture item response theory (IRT) models. Ten indices were considered: Akaike's information criterion (AIC), the corrected AIC (AICc), Bayesian information criterion (BIC), consistent AIC (CAIC), Draper's information criterion (DIC), sample size adjusted BIC (SABIC), relative entropy, the integrated classification likelihood criterion (ICL-BIC), the adjusted Lo-Mendell-Rubin (LMR), and Vuong-Lo-Mendell-Rubin (VLMR). The accuracy of the fit indices was assessed for correct detection of the number of latent classes for different simulation conditions including sample size (2,500 and 5,000), test length (15, 30, and 45), mixture proportions (equal and unequal), number of latent classes (2, 3, and 4), and latent class separation (no-separation and small separation). Simulation study results indicated that as the number of examinees or number of items increased, correct identification rates also increased for most of the indices. Correct identification rates by the different fit indices, however, decreased as the number of estimated latent classes or parameters (i.e., model complexity) increased. Results were good for BIC, CAIC, DIC, SABIC, ICL-BIC, LMR, and VLMR, and the relative entropy index tended to select correct models most of the time. Consistent with previous studies, AIC and AICc showed poor performance. Most of these indices had limited utility for three-class and four-class mixture 3PL model conditions.

In recent years, some models for binary and graded format responses have been proposed to assess unipolar variables or "quasi-traits." These studies have mainly focused on clinical variables that have traditionally been treated as bipolar traits. In the present study, we have made a proposal for unipolar traits measured with continuous response items. The proposed log-logistic continuous unipolar model (LL-C) is remarkably simple and is more similar to the original binary formulation than the graded extensions, which is an advantage. Furthermore, considering that irrational, extreme, or polarizing beliefs could be another domain of unipolar variables, we have applied this proposal to an empirical example of superstitious beliefs. The results suggest that, in certain cases, the standard linear model can be a good approximation to the LL-C model in terms of parameter estimation and goodness of fit, but not trait estimates and their accuracy. The results also show the importance of considering the unipolar nature of this kind of trait when predicting criterion variables, since the validity results were clearly different.

Confirmatory factor analyses (CFA) are often used in psychological research when developing measurement models for psychological constructs. Evaluating CFA model fit can be quite challenging, as tests for exact model fit may focus on negligible deviances, while fit indices cannot be interpreted absolutely without specifying thresholds or cutoffs. In this study, we review how model fit in CFA is evaluated in psychological research using fit indices and compare the reported values with established cutoff rules. For this, we collected data on all CFA models in "Psychological Assessment" from the years 2015 to 2020 (N[subscript Studies]=221). In addition, we reevaluate model fit with newly developed methods that derive fit index cutoffs that are tailored to the respective measurement model and the data characteristics at hand. The results of our review indicate that the model fit in many studies has to be seen critically, especially with regard to the usually imposed independent clusters constraints. In addition, many studies do not fully report all results that are necessary to re-evaluate model fit. We discuss these findings against new developments in model fit evaluation and methods for specification search.

Coefficient omega indices are model-based composite reliability estimates that have become increasingly popular. A coefficient omega index estimates how reliably an observed composite score measures a target construct as represented by a factor in a factor-analysis model; as such, the accuracy of omega estimates is likely to depend on correct model specification. The current paper presents a simulation study to investigate the performance of omega-unidimensional (based on the parameters of a one-factor model) and omega-hierarchical (based on a bifactor model) under correct and incorrect model misspecification for high and low reliability composites and different scale lengths. Our results show that coefficient omega estimates are unbiased when calculated from the parameter estimates of a properly specified model. However, omega-unidimensional produced positively biased estimates when the population model was characterized by unmodeled error correlations or multidimensionality, whereas omega-hierarchical was only slightly biased when the population model was either a one-factor model with correlated errors or a higher-order model. These biases were higher when population reliability was lower and increased with scale length. Researchers should carefully evaluate the feasibility of a one-factor model before estimating and reporting omega-unidimensional.

Multiple imputation (MI) is one of the recommended techniques for handling missing data in ordinal factor analysis models. However, methods for computing MI-based fit indices under ordinal factor analysis models have yet to be developed. In this short note, we introduced the methods of using the standardized root mean squared residual (SRMR) and the root mean square error of approximation (RMSEA) to assess the fit of ordinal factor analysis models with multiply imputed data. Specifically, we described the procedure for computing the MI-based sample estimates and constructing the confidence intervals. Simulation results showed that the proposed methods could yield sufficiently accurate point and interval estimates for both SRMR and RMSEA, especially in conditions with larger sample sizes, less missing data, more response categories, and higher degrees of misfit. Based on the findings, implications and recommendations were discussed.

Accounting for model misspecification in Bayesian structural equation models is an active area of research. We present a uniquely Bayesian approach to misspecification that models the degree of misspecification as a parameter--a parameter akin to the correlation root mean squared residual. The misspecification parameter can be interpreted on its own terms as a measure of absolute model fit and allows for comparing different models fit to the same data. By estimating the degree of misspecification simultaneously with structural parameters, the uncertainty about structural parameters reflects the degree of model misspecification. This results in a model that produces more reliable inference than extant Bayesian structural equation modeling. In addition, the approach estimates the residual covariance matrix that can be the basis for diagnosing misspecifications and updating a hypothesized model. These features are confirmed using simulation studies. Demonstrations with a variety of real-world examples show additional properties of the approach.

Social desirability bias (SDB) has been a major concern in educational and psychological assessments when measuring latent variables because it has the potential to introduce measurement error and bias in assessments. Person-fit indices can detect bias in the form of misfitted response vectors. The objective of this study was to compare the performance of 14 person-fit indices to identify SDB in simulated responses. The area under the curve (AUC) of receiver operating characteristic (ROC) curve analysis was computed to evaluate the predictive power of these statistics. The findings showed that the agreement statistic (A) outperformed all other person-fit indices, while the disagreement statistic (D), dependability statistic (E), and the number of Guttman errors (G) also demonstrated high AUCs to detect SDB. Recommendations for practitioners to use these fit indices are provided.

The present paper introduces a general multidimensional model to measure individual differences in learning within a single administration of a test. Learning is assumed to result from practicing the operations involved in solving the items. The model accounts for the possibility that the ability to learn may manifest differently for correct and incorrect responses, which allows for distinguishing different types of learning effects in the data. Model estimation and evaluation is based on a Bayesian framework. A simulation study is presented that examines the performance of the estimation and evaluation methods. The results show accuracy in parameter recovery as well as good performance in model evaluation and selection. An empirical study illustrates the applicability of the model to data from a logical ability test.

Viable methods for the identification of item misfit or Differential Item Functioning (DIF) are central to scale construction and sound measurement. Many approaches rely on the derivation of a limiting distribution under the assumption that a certain model fits the data perfectly. Typical DIF assumptions such as the monotonicity and population independence of item functions are present even in classical test theory but are more explicitly stated when using item response theory or other latent variable models for the assessment of item fit. The work presented here provides a robust approach for DIF detection that does not assume perfect model data fit, but rather uses Tukey's concept of contaminated distributions. The approach uses robust outlier detection to flag items for which adequate model data fit cannot be established.

In the literature of modern psychometric modeling, mostly related to item response theory (IRT), the fit of model is evaluated through known indices, such as X[superscript 2], M2, and root mean square error of approximation (RMSEA) for absolute assessments as well as Akaike information criterion (AIC), consistent AIC (CAIC), and Bayesian information criterion (BIC) for relative comparisons. Recent developments show a merging trend of psychometric and machine learnings, yet there remains a gap in the model fit evaluation, specifically the use of the area under curve (AUC). This study focuses on the behaviors of AUC in fitting IRT models. Rounds of simulations were conducted to investigate AUC's appropriateness (e.g., power and Type I error rate) under various conditions. The results show that AUC possessed certain advantages under certain conditions such as high-dimensional structure with two-parameter logistic (2PL) and some three-parameter logistic (3PL) models, while disadvantages were also obvious when the true model is unidimensional. It cautions researchers about the dangers of using AUC solely in evaluating psychometric models.

The purpose of this study was to evaluate the degree of classification quality in the basic latent class model when covariates are either included or are not included in the model. To accomplish this task, Monte Carlo simulations were conducted in which the results of models with and without a covariate were compared. Based on these simulations, it was determined that models without a covariate better predicted the number of classes. These findings in general supported the use of the popular three-step approach; with its quality of classification determined to be more than 70% under various conditions of covariate effect, sample size, and quality of indicators. In light of these findings, the practical utility of evaluating classification quality is discussed relative to issues that applied researchers need to carefully consider when applying latent class models.

A sample suffers range restriction (RR) when its variance is reduced comparing with its population variance and, in turn, it fails representing such population. If the RR occurs over the latent factor, not directly over the observed variable, the researcher deals with an indirect RR, common when using convenience samples. This work explores how this problem affects different outputs of the factor analysis: multivariate normality (MVN), estimation process, goodness-of-fit, recovery of factor loadings, and reliability. In doing so, a Monte Carlo study was conducted. Data were generated following the linear selective sampling model, simulating tests varying their sample size (N = 200 and 500 cases), test size (J = 6, 12, 18, and 24 items), loading size (L = 0.50, 0.70, and 0.90), and restriction size (from R = 1, 0.90, 0.80, and so on till 0.10 selection ratio). Our results systematically suggest that an interaction between decreasing the loading size and increasing the restriction size affects the MVN assessment, obstructs the estimation process, and leads to an underestimation of the factor loadings and reliability. However, most of the MVN tests and most of the fit indices employed were nonsensitive to the RR problem. We provide some recommendations to applied researchers.

In modeling missing data, the missing data latent variable of the confirmatory factor model accounts for systematic variation associated with missing data so that replacement of what is missing is not required. This study aimed at extending the modeling missing data approach to tetrachoric correlations as input and at exploring the consequences of switching between models with free and fixed factor loadings. In a simulation study, confirmatory factor analysis (CFA) models with and without a missing data latent variable were used for investigating the structure of data with and without missing data. In addition, the numbers of columns of data sets with missing data and the amount of missing data were varied. The root mean square error of approximation (RMSEA) results revealed that an additional missing data latent variable recovered the degree-of-model fit characterizing complete data when tetrachoric correlations served as input while comparative fit index (CFI) results showed overestimation of this degree-of-model fit. Whereas the results for fixed factor loadings were in line with the assumptions of modeling missing data, the other results showed only partial agreement. Therefore, modeling missing data with fixed factor loadings is recommended.

We examined the effect of estimation methods, maximum likelihood (ML), unweighted least squares (ULS), and diagonally weighted least squares (DWLS), on three population SEM (structural equation modeling) fit indices: the root mean square error of approximation (RMSEA), the comparative fit index (CFI), and the standardized root mean square residual (SRMR). We considered different types and levels of misspecification in factor analysis models: misspecified dimensionality, omitting cross-loadings, and ignoring residual correlations. Estimation methods had substantial impacts on the RMSEA and CFI so that different cutoff values need to be employed for different estimators. In contrast, SRMR is robust to the method used to estimate the model parameters. The same criterion can be applied at the population level when using the SRMR to evaluate model fit, regardless of the choice of estimation method.

A Monte Carlo study was conducted to compare the performance of a level-specific (LS) fit evaluation with that of a simultaneous (SI) fit evaluation in multilevel confirmatory factor analysis (MCFA) models. We extended previous studies by examining their performance under MCFA models with different factor structures across levels. In addition, various design factors and interaction effects between intraclass correlation (ICC) and misspecification type (MT) on their performance were considered. The simulation results demonstrate that the LS outperformed the SI in detecting model misspecification at the between-group level even in the MCFA model with different factor structures across levels. Especially, the performance of LS fit indices depended on the ICC, group size (GS), or MT. More specifically, the results are as follows. First, the performance of root mean square error of approximation (RMSEA) was more promising in detecting misspecified between-level models as GS or ICC increased. Second, the effect of ICC on the performance of comparative fit index (CFI) or Tucker-Lewis index (TLI) depended on the MT. Third, the performance of standardized root mean squared residual (SRMR) improved as ICC increased and this pattern was more clear in structure misspecification than in measurement misspecification. Finally, the summary and implications of the results are discussed.

Prior studies investigating the effects of non-normality in structural equation modeling typically induced non-normality in the indicator variables. This procedure neglects the factor analytic structure of the data, which is defined as the sum of latent variables and errors, so it is unclear whether previous results hold if the source of non-normality is considered. We conducted a Monte Carlo simulation manipulating the underlying multivariate distribution to assess the effect of the source of non-normality (latent, error, and marginal conditions with either multivariate normal or non-normal marginal distributions) on different measures of fit (empirical rejection rates for the likelihood-ratio model test statistic, the root mean square error of approximation, the standardized root mean square residual, and the comparative fit index). We considered different estimation methods (maximum likelihood, generalized least squares, and (un)modified asymptotically distribution-free), sample sizes, and the extent of non-normality in correctly specified and misspecified models to investigate their performance. The results show that all measures of fit were affected by the source of non-normality but with varying patterns for the analyzed estimation methods.

The development of technology-enhanced innovative items calls for practical models that can describe polytomous testlet items. In this study, we evaluate four measurement models that can characterize polytomous items administered in testlets: (a) generalized partial credit model (GPCM), (b) testlet-as-a-polytomous-item model (TPIM), (c) random-effect testlet model (RTM), and (d) fixed-effect testlet model (FTM). Using data from GPCM, FTM, and RTM, we examine performance of the scoring models in multiple aspects: relative model fit, absolute item fit, significance of testlet effects, parameter recovery, and classification accuracy. The empirical analysis suggests that relative performance of the models varies substantially depending on the testlet-effect type, effect size, and trait estimator. When testlets had no or fixed effects, GPCM and FTM led to most desirable measurement outcomes. When testlets had random interaction effects, RTM demonstrated best model fit and yet showed substantially different performance in the trait recovery depending on the estimator. In particular, the advantage of RTM as a scoring model was discernable only when there existed strong random effects and the trait levels were estimated with Bayes priors. In other settings, the simpler models (i.e., GPCM, FTM) performed better or comparably. The study also revealed that polytomous scoring of testlet items has limited prospect as a functional scoring method. Based on the outcomes of the empirical evaluation, we provide practical guidelines for choosing a measurement model for polytomous innovative items that are administered in testlets.

Complex research questions often need large samples to obtain accurate estimates of parameters and adequate power. Combining extant data sets into a large, pooled data set is one way this can be accomplished without expending resources. Measurement invariance (MI) modeling is an established approach to ensure participant scores are on the same scale. There are two major problems when combining independent data sets through MI. First, sample sizes will often be large leading to small differences becoming noninvariant. Second, not all data sets may include the same combination of measures. In this article, we present a method that can deal with both these problems and is user friendly. It is a combination of generating random normal deviates for variables missing completely in combination with assessing model fit using the root mean square error of approximation "good enough principle," based on the hypothesis that the difference between groups is not zero but small. We demonstrate the method by examining MI across eight independent data sets and compare the MI decisions of the traditional and "good enough" approach. Our results show the approach has potential in combining educational data.

Composite reliability, or coefficient omega, can be estimated using structural equation modeling. Composite reliability is usually estimated under the basic independent clusters model of confirmatory factor analysis (ICM-CFA). However, due to the existence of cross-loadings, the model fit of the exploratory structural equation model (ESEM) is often found to be substantially better than that of ICM-CFA. The present study first illustrated the method used to estimate composite reliability under ESEM and then compared the difference between ESEM and ICM-CFA in terms of composite reliability estimation under various indicators per factor, target factor loadings, cross-loadings, and sample sizes. The results showed no apparent difference in using ESEM or ICM-CFA for estimating composite reliability, and the rotation type did not affect the composite reliability estimates generated by ESEM. An empirical example was given as further proof of the results of the simulation studies. Based on the present study, we suggest that if the model fit of ESEM (regardless of the utilized rotation criteria) is acceptable but that of ICM-CFA is not, the composite reliability estimates based on the above two models should be similar. If the target factor loadings are relatively small, researchers should increase the number of indicators per factor or increase the sample size.

Prior research suggests that subscores from a single achievement test seldom add value over a single total score. Such scores typically correspond to subcontent areas in the total content domain, but content subdomains might not provide a sound basis for subscores. Using scores on an inferential reading comprehension test from 625 third, fourth, and fifth graders, two new methods of creating subscores were explored. Three subscores were based on the types of incorrect answers given by students. The fourth was based on temporal efficiency in giving correct answers. All four scores were reliable. The three subscores based on incorrect answers added value and validity. In logistic regression analyses predicting failure to reach proficiency on a statewide test, models including subscores fit better than the model with a single total score. Including the pattern of incorrect responses improved fit in all three grades, whereas including the comprehension efficiency score only modestly improved fit in fourth and fifth grades, but not third grade. Area under the curve (AUC) statistics from receiver operating characteristic (ROC) curves based on the various models were higher for models including subscores than those without subscores. Implications for using models with and without subscores are illustrated and discussed.

The multiple response structure can underlie several different technology-enhanced item types. With the increased use of computer-based testing, multiple response items are becoming more common. This response type holds the potential for being scored polytomously for partial credit. However, there are several possible methods for computing raw scores. This research will evaluate several approaches found in the literature using an approach that evaluates how the inclusion of scoring related to the selection/nonselection of both relevant and irrelevant information is incorporated extending Wilson's approach. Results indicated all methods have potential, but the plus/minus and true/false methods seemed the most promising for items using the "select all that apply" instruction set. Additionally, these methods showed a large increase in information per time unit over the dichotomous method.

Despite the existence of many methods for determining the number of factors, none outperforms the others under every condition. This study compares traditional parallel analysis (TPA), revised parallel analysis (RPA), Kaiser's rule, minimum average partial, sequential X[superscript 2], and sequential root mean square error of approximation, comparative fit index, and Tucker-Lewis index under a realistic scenario in behavioral studies, where researchers employ a closing-fitting parsimonious model with "K" factors to approximate a population model, leading to a trivial model-data misfit. Results show that while traditional and RPA both stand out when zero population-level misfits exist, the accuracy of RPA substantially deteriorates when a "K"-factor model can closely approximate the population. TPA is the least sensitive to trivial misfits and results in the highest accuracy across most simulation conditions. This study suggests the use of TPA for the investigated models. Results also imply that RPA requires further revision to accommodate a degree of model-data misfit that can be tolerated.

In low-stakes assessments, some students may not reach the end of the test and leave some items unanswered due to various reasons (e.g., lack of test-taking motivation, poor time management, and test speededness). Not-reached items are often treated as incorrect or not-administered in the scoring process. However, when the proportion of not-reached items is high, these traditional approaches may yield biased scores and thereby threatening the validity of test results. In this study, we propose a polytomous scoring approach for handling not-reached items and compare its performance with those of the traditional scoring approaches. Real data from a low-stakes math assessment administered to second and third graders were used. The assessment consisted of 40 short-answer items focusing on addition and subtraction. The students were instructed to answer as many items as possible within 5 minutes. Using the traditional scoring approaches, students' responses for not-reached items were treated as either not-administered or incorrect in the scoring process. With the proposed scoring approach, students' nonmissing responses were scored polytomously based on how accurately and rapidly they responded to the items to reduce the impact of not-reached items on ability estimation. The traditional and polytomous scoring approaches were compared based on several evaluation criteria, such as model fit indices, test information function, and bias. The results indicated that the polytomous scoring approaches outperformed the traditional approaches. The complete case simulation corroborated our empirical findings that the scoring approach in which nonmissing items were scored polytomously and not-reached items were considered not-administered performed the best. Implications of the polytomous scoring approach for low-stakes assessments were discussed.

This study examined the impact of omitting covariates interaction effect on parameter estimates in multilevel multiple-indicator multiple-cause models as well as the sensitivity of fit indices to model misspecification when the between-level, within-level, or cross-level interaction effect was left out in the models. The parameter estimates produced in the correct and the misspecified models were compared under varying conditions of cluster number, cluster size, intraclass correlation, and the magnitude of the interaction effect in the population model. Results showed that the two main effects were overestimated by approximately half of the size of the interaction effect, and the between-level factor mean was underestimated. None of comparative fit index, Tucker-Lewis index, root mean square error of approximation, and standardized root mean square residual was sensitive to the omission of the interaction effect. The sensitivity of information criteria varied depending majorly on the magnitude of the omitted interaction, as well as the location of the interaction (i.e., at the between level, within level, or cross level). Implications and recommendations based on the findings were discussed.

The frequent practice of overall fit evaluation for latent variable models in educational and behavioral research is reconsidered. It is argued that since overall plausibility does not imply local plausibility and is only necessary for the latter, local misfit should be considered a sufficient condition for model rejection, even in the case of omnibus model tenability. The argument is exemplified with a comparison of the widely used one-parameter and two-parameter logistic models. A theoretically and practically relevant setting illustrates how discounting local fit and concentrating instead on overall model fit may lead to incorrect model selection, even if a popular information criterion is also employed. The article concludes with the recommendation for routine examination of particular parameter constraints within latent variable models as part of their fit evaluation.

Evaluations of measurement invariance provide essential construct validity evidence--a prerequisite for seeking meaning in psychological and educational research and ensuring fair testing procedures in high-stakes settings. However, the quality of such evidence is partly dependent on the validity of the resulting statistical conclusions. Type I or Type II errors can render measurement invariance conclusions meaningless. The present study used Monte Carlo simulation methods to compare the effects of multiple model parameterizations (linear factor model, Tobit factor model, and categorical factor model) and estimators (maximum likelihood [ML], robust maximum likelihood [MLR], and weighted least squares mean and variance-adjusted [WLSMV]) on the performance of the chi-square test for the exact-fit hypothesis and chi-square and likelihood ratio difference tests for the equal-fit hypothesis for evaluating measurement invariance with ordered polytomous data. The test statistics were examined under multiple generation conditions that varied according to the degree of metric noninvariance, the size of the sample, the magnitude of the factor loadings, and the distribution of the observed item responses. The categorical factor model with WLSMV estimation performed best for evaluating overall model fit, and the categorical factor model with ML and MLR estimation performed best for evaluating change in fit. Results from this study should be used to inform the modeling decisions of applied researchers. However, no single analysis combination can be recommended for all situations. Therefore, it is essential that researchers consider the context and purpose of their study.

S-X[superscript 2] is a popular item fit index that is available in commercial software packages such as "flex"MIRT. However, no research has systematically examined the performance of S-X[superscript 2] for detecting item misfit within the context of the multidimensional graded response model (MGRM). The primary goal of this study was to evaluate the performance of S-X[superscript 2] under two practical misfit scenarios: first, all items are misfitting due to model misspecification, and second, a small subset of items violate the underlying assumptions of the MGRM. Simulation studies showed that caution should be exercised when reporting item fit results of polytomous items using S-X[superscript 2] within the context of the MGRM, because of its inflated false positive rates (FPRs), especially with a small sample size and a long test. S-X[superscript 2] performed well when detecting overall model misfit as well as item misfit for a small subset of items when the ordinality assumption was violated. However, under a number of conditions of model misspecification or items violating the homogeneous discrimination assumption, even though true positive rates (TPRs) of S-X[superscript 2] were high when a small sample size was coupled with a long test, the inflated FPRs were generally directly related to increasing TPRs. There was also a suggestion that performance of S-X[superscript 2] was affected by the magnitude of misfit within an item. There was no evidence that FPRs for fitting items were exacerbated by the presence of a small percentage of misfitting items among them.

Model fit indices are being increasingly recommended and used to select the number of factors in an exploratory factor analysis. Growing evidence suggests that the recommended cutoff values for common model fit indices are not appropriate for use in an exploratory factor analysis context. A particularly prominent problem in scale evaluation is the ubiquity of correlated residuals and imperfect model specification. Our research focuses on a scale evaluation context and the performance of four standard model fit indices: root mean square error of approximate (RMSEA), standardized root mean square residual (SRMR), comparative fit index (CFI), and Tucker-Lewis index (TLI), and two equivalence test-based model fit indices: RMSEAt and CFIt. We use Monte Carlo simulation to generate and analyze data based on a substantive example using the positive and negative affective schedule (N = 1,000). We systematically vary the number and magnitude of correlated residuals as well as nonspecific misspecification, to evaluate the impact on model fit indices in fitting a two-factor exploratory factor analysis. Our results show that all fit indices, except SRMR, are overly sensitive to correlated residuals and nonspecific error, resulting in solutions that are overfactored. SRMR performed well, consistently selecting the correct number of factors; however, previous research suggests it does not perform well with categorical data. In general, we do not recommend using model fit indices to select number of factors in a scale evaluation framework.

The interpretation of psychometric test results is usually based on norm scores. We compared semiparametric continuous norming (SPCN) with conventional norming methods by simulating results for test scales with different item numbers and difficulties via an item response theory approach. Subsequently, we modeled the norm scores based on random samples with varying sizes either with a conventional ranking procedure or SPCN. The norms were then cross-validated by using an entirely representative sample of N = 840,000 for which different measures of norming error were computed. This process was repeated 90,000 times. Both approaches benefitted from an increase in sample size, with SPCN reaching optimal results with much smaller samples. Conventional norming performed worse on data fit, age-related errors, and number of missings in the norm tables. The data fit in conventional norming of fixed subsample sizes varied with the granularity of the age brackets, calling into question general recommendations for sample sizes in test norming. We recommend that test norms should be based on statistical models of the raw score distributions instead of simply compiling norm tables via conventional ranking procedures.

We examine the accuracy of p values obtained using the asymptotic mean and variance (MV) correction to the distribution of the sample standardized root mean squared residual (SRMR) proposed by Maydeu-Olivares to assess the exact fit of SEM models. In a simulation study, we found that under normality, the MV-corrected SRMR statistic provides reasonably accurate Type I errors even in small samples and for large models, clearly outperforming the current standard, that is, the likelihood ratio (LR) test. When data shows excess kurtosis, MV-corrected SRMR p values are only accurate in small models (p = 10), or in medium-sized models (p = 30) if no skewness is present and sample sizes are at least 500. Overall, when data are not normal, the MV-corrected LR test seems to outperform the MV-corrected SRMR. We elaborate on these findings by showing that the asymptotic approximation to the mean of the SRMR sampling distribution is quite accurate, while the asymptotic approximation to the standard deviation is not.

Factor mixture modeling (FMM) has been increasingly used to investigate unobserved population heterogeneity. This study examined the issue of covariate effects with FMM in the context of measurement invariance testing. Specifically, the impact of excluding and misspecifying covariate effects on measurement invariance testing and class enumeration was investigated via Monte Carlo simulations. Data were generated based on FMM models with (1) a zero covariate effect, (2) a covariate effect on the latent class variable, and (3) covariate effects on both the latent class variable and the factor. For each population model, different analysis models that excluded or misspecified covariate effects were fitted. Results highlighted the importance of including proper covariates in measurement invariance testing and evidenced the utility of a model comparison approach in searching for the correct specification of covariate effects and the level of measurement invariance. This approach was demonstrated using an empirical data set. Implications for methodological and applied research are discussed.

Bayesian structural equation modeling (BSEM) is a flexible tool for the exploration and estimation of sparse factor loading structures; that is, most cross-loading entries are zero and only a few important cross-loadings are nonzero. The current investigation was focused on the BSEM with small-variance normal distribution priors (BSEM-N) for both variable selection and model estimation. The prior sensitivity in BSEM-N was explored in factor analysis models with sparse loading structures through a simulation study (Study 1) and an empirical example (Study 2). Study 1 examined the prior sensitivity in BSEM-N based on the model fit, population model recovery, true and false positive rates, and parameter estimation. Seven shrinkage priors on cross-loadings and five noninformative/vague priors on other model parameters were examined. Study 2 provided a real data example to illustrate the impact of various priors on model fit and parameter selection and estimation. Results indicated that when the 95% credible intervals of shrinkage priors barely covered the population cross-loading values, it resulted in the best balance between true and false positives. If the goal is to perform variable selection, a sparse cross-loading structure is required, preferably with a minimal number of nontrivial cross-loadings and relatively high primary loading values. To improve parameter estimates, a relatively large prior variance is preferred. When cross-loadings are relatively large, BSEM-N with zero-mean priors is not recommended for the estimation of cross-loadings and factor correlations.

This study compares automated methods to develop short forms of psychometric scales. Obtaining a short form that has both adequate internal structure and strong validity with respect to relationships with other variables is difficult with traditional methods of short-form development. Metaheuristic algorithms can select items for short forms while optimizing on several validity criteria, such as adequate model fit, composite reliability, and relationship to external variables. Using a Monte Carlo simulation study, this study compared existing implementations of the ant colony optimization, Tabu search, and genetic algorithm to select short forms of scales, as well as a new implementation of the simulated annealing algorithm. Selection of short forms of scales with unidimensional, multidimensional, and bifactor structure were evaluated, with and without model misspecification and/or an external variable. The results showed that when the confirmatory factor analysis model of the full form of the scale was correctly specified or had only minor misspecification, the four algorithms produced short forms with good psychometric qualities that maintained the desired factor structure of the full scale. Major model misspecification resulted in worse performance for all algorithms, but including an external variable only had minor effects on results. The simulated annealing algorithm showed the best overall performance as well as robustness to model misspecification, while the genetic algorithm produced short forms with worse fit than the other algorithms under conditions with model misspecification.

The main thesis of the present study is to use the Bayesian structural equation modeling (BSEM) methodology of establishing approximate measurement invariance (A-MI) using data from a national examination in Saudi Arabia as an alternative to not meeting strong invariance criteria. Instead, we illustrate how to account for the absence of measurement invariance using relative compared to exact criteria. A secondary goal was to compare latent means across groups using invariant parameters only and through utilizing exact and relative evaluative-MI protocol suggested equivalence of the thresholds using prior variances equal to 0.10. Subsequent differences between groups were evaluated using effect size criteria and the prior-posterior predictive p-value (PPPP), which proved to be invaluable in attesting for differences that are beyond zero, some meaningless nonzero estimate, and the three commonly used indices of effect sizes described by Cohen in 1988 (i.e., 0.20, 0.50, and 0.80). Results substantiated the use of the PPPP for evaluating mean differences across groups when utilizing nonexact evaluative criteria.

Exploratory factor analysis (EFA) is widely used by researchers in the social sciences to characterize the latent structure underlying a set of observed indicator variables. One of the primary issues that must be resolved when conducting an EFA is determination of the number of factors to retain. There exist a large number of statistical tools designed to address this question, with none being universally optimal across applications. Recently, researchers have investigated the use of model fit indices that are commonly used in the conduct of confirmatory factor analysis to determine the number of factors to retain in EFA. These results have yielded mixed results, appearing to be effective when used in conjunction with normally distributed indicators, but not being as effective for categorical indicators. The purpose of this simulation study was to compare the performance of difference values for several fit indices as a method for identifying the optimal number of factors to retain in an EFA, with parallel analysis, which is one of the most reliable such extant methods. Results of the simulation demonstrated that the use of fit index difference values outperformed parallel analysis for categorical indicators, and for normally distributed indicators when factor loadings were small. Implications of these findings are discussed.

This study compares two missing data procedures in the context of ordinal factor analysis models: pairwise deletion (PD; the default setting in Mplus) and multiple imputation (MI). We examine which procedure demonstrates parameter estimates and model fit indices closer to those of complete data. The performance of PD and MI are compared under a wide range of conditions, including number of response categories, sample size, percent of missingness, and degree of model misfit. Results indicate that both PD and MI yield parameter estimates similar to those from analysis of complete data under conditions where the data are missing completely at random (MCAR). When the data are missing at random (MAR), PD parameter estimates are shown to be severely biased across parameter combinations in the study. When the percentage of missingness is less than 50%, MI yields parameter estimates that are similar to results from complete data. However, the fit indices (i.e., X[superscript 2], RMSEA, and WRMR) yield estimates that suggested a worse fit than results observed in complete data. We recommend that applied researchers use MI when fitting ordinal factor models with missing data. We further recommend interpreting model fit based on the TLI and CFI incremental fit indices.

This study presents new models for item response functions (IRFs) in the framework of the D-scoring method (DSM) that is gaining attention in the field of educational and psychological measurement and largescale assessments. In a previous work on DSM, the IRFs of binary items were estimated using a logistic regression model (LRM). However, the LRM underestimates the item true scores at the top end of the D-scale (ranging from 0 to 1), especially for relatively difficult items. This entails underestimation of true D-scores, inaccuracy in the estimates of their standard errors, and other psychometric issues. The inverse-regression adjustments used to fix this problem are too complicated for regular applications of the DSM and not in line with its simplicity. This issue is resolved with the IRF models proposed in this study, referred to as rational function models (RFMs) with one parameter (RFM1), two parameters (RFM2), and three parameters (RFM3). The proposed RFMs are discussed and illustrated with simulated and real data.

Factor score regression (FSR) is a popular alternative for structural equation modeling. Naively applying FSR induces bias for the estimators of the regression coefficients. Croon proposed a method to correct for this bias. Next to estimating effects without bias, interest often lies in inference of regression coefficients or in the fit of the model. In this article, we propose fit indices for FSR that can be used to inspect the model fit. We also introduce a model comparison test based on one of these newly proposed fit indices that can be used for inference of the estimators on the regression coefficients. In a simulation study we compare FSR with Croon's corrections and structural equation modeling in terms of bias of the regression coefficients, Type I error rate and power.

Item response theory (IRT) is a statistical paradigm for developing educational tests and assessing students. IRT, however, currently lacks an established graphical method for examining model fit for the three-parameter logistic model, the most flexible and popular IRT model in educational testing. A method is presented here to do this. The graph, which is referred to herein as a "bin plot," is the IRT equivalent of a scatterplot for linear regression. Bin plots display a conventional IRT trace line (with ability on the horizontal axis and probability correct on the vertical axis). Students are binned according to how well they performed on the entire test, and the proportion of students in each bin who answered the focal question correctly is displayed on the graph as points above or below the trace line. With this arrangement, the difference between each point and the trace line is the residual for the bin. Confidence intervals can be added to the observed proportions in order to display uncertainty. Computer simulations were used to test four alternative ways for binning students. These simulations showed that binning students according to number of questions they answered correctly on the entire test works best. Simulations also showed confidence intervals for bin plots had coverage probabilities close to nominal values for common testing scenarios, but that there are scenarios in which confidence intervals had inflated error rates.

A simulation study was conducted to investigate the model size effect when confirmatory factor analysis (CFA) models include many ordinal items. CFA models including between 15 and 120 ordinal items were analyzed with mean- and variance-adjusted weighted least squares to determine how varying sample size, number of ordered categories, and misspecification affect parameter estimates, standard errors of parameter estimates, and selected fit indices. As the number of items increased, the number of admissible solutions and accuracy of parameter estimates improved, even when models were misspecified. Also, standard errors of parameter estimates were closer to empirical standard deviation values as the number of items increased. When evaluating goodness-of-fit for ordinal CFA with many observed indicators, researchers should be cautious in interpreting the root mean square error of approximation, as this value appeared overly optimistic under misspecified conditions.

This study investigated the effect the number of observed variables (p) has on three structural equation modeling indices: the comparative fit index (CFI), the Tucker--Lewis index (TLI), and the root mean square error of approximation (RMSEA). The behaviors of the population fit indices and their sample estimates were compared under various conditions created by manipulating the number of observed variables, the types of model misspecification, the sample size, and the magnitude of factor loadings. The results showed that the effect of p on the population CFI and TLI depended on the type of specification error, whereas a higher "p" was associated with lower values of the population RMSEA regardless of the type of model misspecification. In finite samples, all three fit indices tended to yield estimates that suggested a worse fit than their population counterparts, which was more pronounced with a smaller sample size, higher "p," and lower factor loading.

Measures initially designed to be single-trait often yield data that are compatible with both an essentially unidimensional factor-analysis (FA) solution and a correlated-factors solution. For these cases, this article proposes an approach aimed at providing information for deciding which of the two solutions is the most appropriate and useful. The procedures we propose are an FA extension of the "added-value" procedures initially proposed for subscale scores in educational testing. The basic principle is that the multiple FA solution is defensible when the factor score estimates of the primary factors are better measures of these factors than score estimates derived from a unidimensional or second-order solution. Methodologically, new results are obtained, and relations with factor indeterminacy measures and second-order FA are discussed. The procedures have been implemented in a noncommercial and widely known program for exploratory FA. The functioning of the proposal is assessed by means of a simulation study, and its usefulness is illustrated with a real-data example in the personality domain.

One of the most cited methodological issues is with the response format, which is traditionally a single-response Likert response format. Therefore, our study aims to elucidate and illustrate an alternative response format and analytic technique, Thurstonian item response theory (IRT), for analyzing data from surveys using an alternate response format, the forced-choice format. Specifically, we strove to give a thorough introduction of Thurstonian IRT at a more elementary level than previous publications in order to widen the possible audience. This article presents analyses and comparison of two versions of a self-report scale, one version using a single-response format and the other using a forced-choice format. Drawing from lessons learned from our study and literature, we present a number of recommendations for conducting research using the forced-choice format and Thurstonian IRT, as well as suggested avenues for future research.

The article reports three simulation studies conducted to find out whether the effect of a time limit for testing impairs model fit in investigations of structural validity, whether the representation of the assumed source of the effect prevents impairment of model fit and whether it is possible to identify and discriminate this method effect from another method effect. Omissions due to the time limit for testing were not considered as missing data but as information on the participants' processing speed. In simulated data the presence of a time-limit effect impaired comparative fit index and nonnormed fit index whereas normed chi-square, root mean square error of approximation, and standardized root mean square residual indicated good model fit. The explicit consideration of the effect due to the time limit by an additional component of the model improved model fit. Effect-specific assumptions included in the model of measurement enabled the discrimination of the effect due to the time limit from another possible method effect.

Studies comparing groups that are at different levels of multilevel data (namely, cross-level groups) using the same measure are not unusual such as student and teacher agreement in education or congruence between patient and physician perceptions in health research. Although establishing measurement invariance (MI) between these groups is important, testing MI is methodologically challenging because the groups compared for MI are at different levels with one group nested within the other group. We propose a multilevel confirmatory factor analysis (CFA) model that allows MI testing between cross-level groups at the between level and demonstrated testing MI between students and teachers using the promoting social interaction scale. Along with the demonstration, some methodological issues in implementing the proposed model (e.g., cluster invariance and reliability) and evaluating the model fit of multilevel CFA (e.g., [delta]CFI and level-specific fit indices) and alternative approaches to the proposed model are discussed.

Bayesian confirmatory factor analysis (CFA) offers an alternative to frequentist CFA based on, for example, maximum likelihood estimation for the assessment of reliability and validity of educational and psychological measures. For increasing sample sizes, however, the applicability of current fit statistics evaluating model fit within Bayesian CFA is limited. We propose, therefore, a Bayesian variant of the root mean square error of approximation (RMSEA), the BRMSEA. A simulation study was performed with variations in model misspecification, factor loading magnitude, number of indicators, number of factors, and sample size. This showed that the 90% posterior probability interval of the BRMSEA is valid for evaluating model fit in large samples (N= 1,000), using cutoff values for the lower (<0.05) and upper limit (<0.08) as guideline. An empirical illustration further shows the advantage of the BRMSEA in large sample Bayesian CFA models. In conclusion, it can be stated that the BRMSEA is well suited to evaluate model fit in large sample Bayesian CFA models by taking sample size and model complexity into account.

Attribute structure is an explicit way of presenting the relationship between attributes in diagnostic measurement. The specification of attribute structures directly affects the classification accuracy resulted from psychometric modeling. This study provides a conceptual framework for understanding misspecifications of attribute structures. Under the framework, each attribute structure can be represented through an external shape and an internal organization. A simulation study and an application example were used to investigate how misspecification of external shapes and internal organizations affects model fit and item fit assessments, and respondent classification. The proposed framework and simulation results aim to support using attribute structures to (a) develop better diagnostic assessments and (b) inform theories of constructs.

Previous research includes frequent admonitions regarding the importance of establishing connectivity in data collection designs prior to the application of Rasch models. However, details regarding the influence of characteristics of the linking sets used to establish connections among facets, such as locations on the latent variable, model-data fit, and sample size, have not been thoroughly explored. These considerations are particularly important in assessment systems that involve large proportions of missing data (i.e., sparse designs) and are associated with high-stakes decisions, such as teacher evaluations based on teaching observations. The purpose of this study is to explore the influence of characteristics of linking sets in sparsely connected rating designs on examinee, rater, and task estimates. A simulation design whose characteristics were intended to reflect practical large-scale assessment networks with sparse connections were used to consider the influence of locations on the latent variable, model-data fit, and sample size within linking sets on the stability and model-data fit of estimates. Results suggested that parameter estimates for examinee and task facets are quite robust to modifications in the size, model-data fit, and latent-variable location of the link. Parameter estimates for the rater, while still quite robust, are more sensitive to reductions in link size. The implications are discussed as they relate to research, theory, and practice.

It is common practice for assessment programs to organize qualifying sessions during which the raters (often known as "markers" or "judges") demonstrate their consistency before operational rating commences. Because of the high-stakes nature of many rating activities, the research community tends to continuously explore new methods to analyze rating data. We used simulated and empirical data from two high-stakes language assessments, to propose a new approach, based on social network analysis and exponential graph models, to evaluate the readiness of a group of raters for operational rating. The results of this innovative approach are compared with the results of a Rasch analysis, which is a well-established approach for the analysis of such data. We also demonstrate how the new approach can be practically used to investigate important research questions such as whether rater severity is stable across rating tasks. The merits of the new approach, and the consequences for practice are discussed.